Can GPU drivers cause a total system shutdown?

[roy_g_biv]

Hey all,

As per the title really - I'm getting some really rough crashes and I've narrowed it down to my GPU. I'll put some details below for anyone who's interested, but the basic question is whether this is something that's even possible.

Can a GPU driver cause the entire system to shut down, or is this by necessity a hardware problem?

----
I've had an RX 480 and Ryzen 3600, slowly building my new PC (Ryzen 5900x, aiming for an RTX 3080 or 6800XT).

Everything was working fine, games running etc.

I get a hold of the 5900x first, and a new case. Install the CPU, move the mobo etc. across. Everything running fine, BIOS latest version, chipset drivers installed.

2 days of normal use (browsing, videos etc.) go by uneventfully, until I get a hankering to play a game.

Fire it up, crashes completely after a few seconds of showing a 3D model in the menu. Completely off, no lights on the tower.

Turn back on, run Furmark, same deal - runs for a couple of seconds, PC dies.

I update drivers on the AMD website, same deal.

Factory reset AMD drivers, same deal.

Test memory, stress test CPU, check all connections, everything fine.

I swap in an old Radeon HD7770, and that runs Furmark fine. I figure I can simulate the extra wattage of the 480 by running a concurrent CPU stress test on Prime95, and everything is stable and fine. Presumably this eliminates the previous most likely cause of 'PSU overloaded'

PSU is a Seasonic Prime Platinum 750W. Bought a month or two ago.


So it seems to point to a GPU hardware/software problem right?


It just seems really strange for that kind of crash to occur from a GPU.

I know I'm getting a new GPU, but I'd be giving my RX480 to a friend so I'd like to resolve the issue if possible. Or at least find out ealy if I have to replace my PSU.


Anyone have any ideas how I'd go about isolating the problem further, and hopefully fixing?


Thanks!

 

[LtMatt]

The drivers no, but a faulty GPU or PSU could cause that.

I am assuming you have tested with BIOS settings at optimised default and using the latest BIOS first...

As the PSU is new, i I would firstly try the RX 480 in a different system if you can to see if similar happens.

If the GPU checks out as okay, look at the PSU/cables.

 

[roy_g_biv]

Ah interesting, I didn't think to check the integrity of the cables. It's a modular so I can swap in one of the others this evening. Thanks
Testing on another tower will take a little while, but I'll give it a go!

Since everything was running fine on the same PSU before the case change, I'm thinking maybe in the changeover process something got knocked loose in the graphics card and broke it. No idea how to even diagnose physical issues though

 

[mattyfc]

If system is shutting off totally and you have to toggle the switch on the PSU to get it going again, could well be OCP or some other protection kicking in on the PSU.

Draw could be too high on that rail or transient power spike, though your model of Seasonic PSU shouldn't be effected by that issue normally.

There have been issues with RX 480 before drawing too much power from the board in the past:

https://www.extremetech.com/extreme...-will-release-updated-driver-by-late-thursday

Maybe combination of RX480, 5900x is pulling too much watts and amps from the board and tripping PSU OCP.

 

[roy_g_biv]

If system is shutting off totally and you have to toggle the switch on the PSU to get it going again, could well be OCP or some other protection kicking in on the PSU.

Draw could be too high on that rail or transient power spike, though your model of Seasonic PSU shouldn't be effected by that issue normally.

There have been issues with RX 480 before drawing too much power from the board in the past:

https://www.extremetech.com/extreme...-will-release-updated-driver-by-late-thursday

Maybe combination of RX480, 5900x is pulling too much watts and amps from the board and tripping PSU OCP.

Hmm, what do you mean by 'toggle the switch on the PSU'? I was able to just use the main PC on/off switch to turn it back on, didn't need to reach into the back to do anything.

I actually have the ROG Strix RX 480, which uses an 8-pin PCI-E power connector, I don't think the issues were present in those (the article talks about bringing the 6-pin out of spec)

 

[mattyfc]

I would try to isolate the issue, try rx480 in different machine as mentioned above, swap in your old cpu and try find which component or combination is causing it.

Not exactly sure how seasonic ocp protection works. If that is what is causing the problem.

I had similar OCP shutdown issue with an EVGA psu recently and toggling switch only way to get it on again after.

 

[roy_g_biv]

If anyone finds this in future: I managed to figure out the cause

Looks like reseating the GPU during the installation of the new parts had compromised the contact of the thermal paste/pads.

I ran Furmark straight from a cold boot, and that gave me enough time to see the temperature spike to 90C just before shutdown.

Opened it up, cleaned and reapplied thermal paste, closed it back up - totally fine now. Temps steady at 60C at full load

Thanks everyone for your help troubleshooting

 

[LtMatt]

If anyone finds this in future: I managed to figure out the cause

Looks like reseating the GPU during the installation of the new parts had compromised the contact of the thermal paste/pads.

I ran Furmark straight from a cold boot, and that gave me enough time to see the temperature spike to 90C just before shutdown.

Opened it up, cleaned and reapplied thermal paste, closed it back up - totally fine now. Temps steady at 60C at full load

Thanks everyone for your help troubleshooting

Glad to hear it.